| Message ID | 20170922183803.10701-10-hdegoede@redhat.com (mailing list archive) |
|---|---|
| State | Awaiting Upstream, archived |
| Headers | show |
Hi Hans, On 2017년 09월 23일 03:37, Hans de Goede wrote: > Cherry Trail SoCs have a built-in USB-role mux for switching between > the host and device controllers, rather then using an external mux > controller by a GPIO. > > There is a driver using the mux-subsys to control this mux, this > commit adds support to the intel-int3496 driver to get a mux_controller > handle for the mux and set the mux through the mux-subsys rather then > through a GPIO. > > Signed-off-by: Hans de Goede <hdegoede@redhat.com> > --- > Changes in v2: > -Drop || COMPILE_TEST from Kconfig depends on, as we will now fail to > compile on !X86 > -Minor code style tweaks > --- > drivers/extcon/Kconfig | 3 +- > drivers/extcon/extcon-intel-int3496.c | 59 +++++++++++++++++++++++++++++++++++ > 2 files changed, 61 insertions(+), 1 deletion(-) Acked-by: Chanwoo Choi <cw00.choi@samsung.com> [snip]
Hi, On 18-10-17 04:33, Chanwoo Choi wrote: > Hi Hans, > > On 2017년 09월 23일 03:37, Hans de Goede wrote: >> Cherry Trail SoCs have a built-in USB-role mux for switching between >> the host and device controllers, rather then using an external mux >> controller by a GPIO. >> >> There is a driver using the mux-subsys to control this mux, this >> commit adds support to the intel-int3496 driver to get a mux_controller >> handle for the mux and set the mux through the mux-subsys rather then >> through a GPIO. >> >> Signed-off-by: Hans de Goede <hdegoede@redhat.com> >> --- >> Changes in v2: >> -Drop || COMPILE_TEST from Kconfig depends on, as we will now fail to >> compile on !X86 >> -Minor code style tweaks >> --- >> drivers/extcon/Kconfig | 3 +- >> drivers/extcon/extcon-intel-int3496.c | 59 +++++++++++++++++++++++++++++++++++ >> 2 files changed, 61 insertions(+), 1 deletion(-) > > Acked-by: Chanwoo Choi <cw00.choi@samsung.com> Note that there have been some comments on this series, and it is not sure yet how we are going to end up handling this. So please do not merge this yet, as we may end up with another solution. Regards, Hans
On 2017년 10월 18일 18:14, Hans de Goede wrote: > Hi, > > On 18-10-17 04:33, Chanwoo Choi wrote: >> Hi Hans, >> >> On 2017년 09월 23일 03:37, Hans de Goede wrote: >>> Cherry Trail SoCs have a built-in USB-role mux for switching between >>> the host and device controllers, rather then using an external mux >>> controller by a GPIO. >>> >>> There is a driver using the mux-subsys to control this mux, this >>> commit adds support to the intel-int3496 driver to get a mux_controller >>> handle for the mux and set the mux through the mux-subsys rather then >>> through a GPIO. >>> >>> Signed-off-by: Hans de Goede <hdegoede@redhat.com> >>> --- >>> Changes in v2: >>> -Drop || COMPILE_TEST from Kconfig depends on, as we will now fail to >>> compile on !X86 >>> -Minor code style tweaks >>> --- >>> drivers/extcon/Kconfig | 3 +- >>> drivers/extcon/extcon-intel-int3496.c | 59 +++++++++++++++++++++++++++++++++++ >>> 2 files changed, 61 insertions(+), 1 deletion(-) >> >> Acked-by: Chanwoo Choi <cw00.choi@samsung.com> > > Note that there have been some comments on this series, and it > is not sure yet how we are going to end up handling this. So > please do not merge this yet, as we may end up with another > solution. Sure. I don't merge only this patch. After finishing the review of this patchset, one maintainer apply all patches and then send immutable pull request.
diff --git a/drivers/extcon/Kconfig b/drivers/extcon/Kconfig index a7bca4207f44..168f9d710ea0 100644 --- a/drivers/extcon/Kconfig +++ b/drivers/extcon/Kconfig @@ -44,7 +44,8 @@ config EXTCON_GPIO config EXTCON_INTEL_INT3496 tristate "Intel INT3496 ACPI device extcon driver" - depends on GPIOLIB && ACPI && (X86 || COMPILE_TEST) + depends on GPIOLIB && ACPI && X86 + select MULTIPLEXER help Say Y here to enable extcon support for USB OTG ports controlled by an Intel INT3496 ACPI device. diff --git a/drivers/extcon/extcon-intel-int3496.c b/drivers/extcon/extcon-intel-int3496.c index 1a45e745717d..3c8e17449c12 100644 --- a/drivers/extcon/extcon-intel-int3496.c +++ b/drivers/extcon/extcon-intel-int3496.c @@ -23,8 +23,13 @@ #include <linux/gpio.h> #include <linux/interrupt.h> #include <linux/module.h> +#include <linux/mux/consumer.h> +#include <linux/mux/usb.h> #include <linux/platform_device.h> +#include <asm/cpu_device_id.h> +#include <asm/intel-family.h> + #define INT3496_GPIO_USB_ID 0 #define INT3496_GPIO_VBUS_EN 1 #define INT3496_GPIO_USB_MUX 2 @@ -37,6 +42,8 @@ struct int3496_data { struct gpio_desc *gpio_usb_id; struct gpio_desc *gpio_vbus_en; struct gpio_desc *gpio_usb_mux; + struct mux_control *usb_mux; + bool usb_mux_set; int usb_id_irq; }; @@ -56,11 +63,32 @@ static const struct acpi_gpio_mapping acpi_int3496_default_gpios[] = { { }, }; +static struct mux_lookup acpi_int3496_cht_mux_lookup[] = { + { + .provider = "intel_cht_usb_mux", + .dev_id = "INT3496:00", + .mux_name = "usb-role-mux", + }, +}; + +#define ICPU(model) { X86_VENDOR_INTEL, 6, model, X86_FEATURE_ANY, } + +static const struct x86_cpu_id cht_cpu_ids[] = { + ICPU(INTEL_FAM6_ATOM_AIRMONT), /* Braswell, Cherry Trail */ + {} +}; + +static bool int3496_soc_has_mux(void) +{ + return x86_match_cpu(cht_cpu_ids); +} + static void int3496_do_usb_id(struct work_struct *work) { struct int3496_data *data = container_of(work, struct int3496_data, work.work); int id = gpiod_get_value_cansleep(data->gpio_usb_id); + int ret; /* id == 1: PERIPHERAL, id == 0: HOST */ dev_dbg(data->dev, "Connected %s cable\n", id ? "PERIPHERAL" : "HOST"); @@ -72,6 +100,22 @@ static void int3496_do_usb_id(struct work_struct *work) if (!IS_ERR(data->gpio_usb_mux)) gpiod_direction_output(data->gpio_usb_mux, id); + if (data->usb_mux) { + /* + * The mux framework expects multiple competing users, we must + * release our previous setting before applying the new one. + */ + if (data->usb_mux_set) + mux_control_deselect(data->usb_mux); + + ret = mux_control_select(data->usb_mux, + id ? MUX_USB_DEVICE : MUX_USB_HOST); + if (ret) + dev_err(data->dev, "Error setting mux: %d\n", ret); + + data->usb_mux_set = ret == 0; + } + if (!IS_ERR(data->gpio_vbus_en)) gpiod_direction_output(data->gpio_vbus_en, !id); @@ -107,6 +151,21 @@ static int int3496_probe(struct platform_device *pdev) data->dev = dev; INIT_DELAYED_WORK(&data->work, int3496_do_usb_id); + if (int3496_soc_has_mux()) { + mux_add_table(acpi_int3496_cht_mux_lookup, + ARRAY_SIZE(acpi_int3496_cht_mux_lookup)); + data->usb_mux = devm_mux_control_get(dev, "usb-role-mux"); + /* Doing this here keeps our error handling clean. */ + mux_remove_table(acpi_int3496_cht_mux_lookup, + ARRAY_SIZE(acpi_int3496_cht_mux_lookup)); + if (IS_ERR(data->usb_mux)) { + ret = PTR_ERR(data->usb_mux); + if (ret != -EPROBE_DEFER) + dev_err(dev, "can't get mux: %d\n", ret); + return ret; + } + } + data->gpio_usb_id = devm_gpiod_get(dev, "id", GPIOD_IN); if (IS_ERR(data->gpio_usb_id)) { ret = PTR_ERR(data->gpio_usb_id);
Cherry Trail SoCs have a built-in USB-role mux for switching between the host and device controllers, rather then using an external mux controller by a GPIO. There is a driver using the mux-subsys to control this mux, this commit adds support to the intel-int3496 driver to get a mux_controller handle for the mux and set the mux through the mux-subsys rather then through a GPIO. Signed-off-by: Hans de Goede <hdegoede@redhat.com> --- Changes in v2: -Drop || COMPILE_TEST from Kconfig depends on, as we will now fail to compile on !X86 -Minor code style tweaks --- drivers/extcon/Kconfig | 3 +- drivers/extcon/extcon-intel-int3496.c | 59 +++++++++++++++++++++++++++++++++++ 2 files changed, 61 insertions(+), 1 deletion(-)